Tech Source, RunTime partner to accelerate GPGPU development with AMD/OpenCL-compliant XMC
NewsOctober 30, 2012
To enhance OpenCL GPGPU development, Tech Source, Inc. has partnered with RunTime Computing Solutions to optimize the use of RunTime’s VSI/Pro GPU math and signal processing library on its Condor 3000 XMC video graphics card. Optimization of the Condor 3000 cards with RunTime’s math and sensor processing library is aimed at easing GPGPU development time and effort.
The partnership tailored RunTime’s VSI/Pro GPU library for operation on the Condor 3000, which now enables GPU developers to easily port existing VSIPL-standard-based programs, access signal processing performance on the video graphics card without OpenCL experience, and expand the Condor 3000 through the addition of outside OpenCL kernels. Designed with GPGPU applications in mind, the Condor 3000 is especially suited for military/aerospace applications like image analysis, video surveillance, radar, and sonar.
Based on AMD’s Radeon E6760 embedded GPUs, Condor 3000 cards are equipped with a peak single precision floating point performance of 576 GFLOPs and contain native support of the OpenCL programming language. Integration of RunTime’s VSI/Pro GPU library also adds real-time or near-real-time performance-portable signal processing capabilities for developers, including convolutions, correlations, Fast Fourier Transforms (FFTs), and linear/vector/matrix operations.
RunTime Computing Solutions President Jennifer Skjellum commented on the partnership, saying “Our library provides Condor 3000 users with this performance-portable library that enables existing VSIPL programs on the new hardware thereby providing low cost portability for existing architectures. Since the VSIPL standard is a proven way - particularly in defense and mil/aero systems – to write high performance signal processing applications, many existing applications will work on the Condor 3000, speeding the ability to use this embedded GPGPU.
